Neural Oscillations and Functional Significances for Prioritizing Dual-Task Walking in Parkinson’s Disease

Participants

Thirty-four people with PD (mean age: 65.0±5.8 years) participated in this study. The inclusion criteria were (1) a diagnosis of idiopathic PD according to the United Kingdom PD Society Brain Bank clinical diagnostic criteria [39], (2) PD onset at age ≥40 years, (3) a Mini-Mental State Examination score > 25, and (4) ability to walk 10 m without a walking aid or help from another person when off medication. Patients were excluded if they had a history of brain surgery (e.g., deep brain stimulation) or other diseases and conditions that could influence balance ability or had moderate or severe postural tremor or kinetic tremor in the hands (to prevent a possible effect on the manual task). Tremor severity was defined by a score > 2 on item 3.15 (postural tremor) or 3.16 (kinetic tremor) in the Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale [40]. Table 1 presents the demographic data and clinical characteristics of the participants. All procedures in this study were approved by the National Taiwan University Hospital Research Ethics Committee (Clinical Trial Registration No.: NCT03298503), and all participants gave written informed consent. The sample size was calculated based on data from a previous study [21], which indicated that 16 participants would be sufficient to detect the effect of task prioritization on dual-task walking (effect size d = 0.87, power = 0.9,α= 0.05).

Experimental conditions and data recording

All clinical assessments and dual-task walking examinations were performed in the morning, at least 12 h after the most recent administration of anti-parkinsonian medications (i.e., medication-off testing). For the dual-task walking, participants were instructed to walk forwardly on an electronic walkway (GAITRite, CIR Systems Inc., Franklin, NJ, USA; sampling rate: 100 Hz) while controlling a pair of inter-locking rings (Fig. 1). To stabilize gait measures with the GAITRite walkway (length: 5.20 m; active area: 4.27 m), the participants walked 2 m before and after the walkway for acceleration and deceleration. For the ring task, the participants flexed elbows at 90°, carefully preventing ring touching while holding two sticks with inter-locking rings (diameter: 4 cm) on the top during walking [20, 37]. The ring task has been proposed as a motor task with relatively low physical demand [37]. If the two rings touched, the computer recorded the event with an A/D card (USB-6221, National Instruments, Austin, TX, USA; sampling rate: 1 kHz). All participants performed the GP condition and MP condition in a random sequence. In the GP condition, the participants were instructed as follows: “The walking task is the primary task. Pay most attention to taking big steps during walking and then to preventing the two rings from touching if possible”. In the MP condition, the participants were instructed as follows: “The ring task is the primary task. Pay most attention to preventing the two rings from touching and then to taking big steps while walking if possible”. To minimize visual feedback from the gait or ring task, the participants were asked to look straight ahead and not at their feet or the rings during walking. Two practice trials were conducted before eight testing trials in each experimental condition. The resting time was 30 s between two testing-trails and was 1 min between the GP and MP conditions for avoiding the effect of muscle fatigue. Immediately following each test trial, the participants were asked to rate the percentage of their attention that they felt had been directed towards “walking with taking big steps” and towards “preventing the two rings from touching”, using an analogue scale (0%–100%) [18].

Fig. 1

Schematic representation of the experimental setup.

Cortical activities were recorded using a NuAmps 32-channel amplifier (NeuroScan Inc., El Paso, TX, USA). The EEG electrodes (Fp1/2, Fz, F3/4, F7/8, FT7/8, FCz, FC3/4, Cz, C3/4, CPz, CP3/4, Pz, P3/4, T3/4, T5/6, TP7/8, Oz, and O1/2) were placed based on the 10–20 electrode system of the International Federation. The ground electrode was placed along the midline, ahead of Fz. During walking, participants carried the EEG amplifier in a backpack with proper fixation as with a mobile EEG system. The wires between the EEG cap and amplifier were fixed on the upper back by several strips of adhesive tape to avoid artifacts caused by the movement of EEG wires. The cable between the amplifier and computer was suspended to avoid interfering with walking. To monitor vertical and horizontal eye movements and blinks, electrodes were placed above the left eyebrow, below the left eye, and horizontally on the outer canthi of both eyes. The impedances of all electrodes were below 5 kΩ, and all electrodes were referenced to linked mastoids of both sides. EEG data were band-pass filtered at 0.1–100 Hz with a 1 kHz sampling rate. All behavioral data and EEG data were synchronized.

Data analysis

At the behavior level, speed (cm/s), cadence (steps/min), step length (cm), and the coefficient of variation (CV, %) of step length were calculated for the walking task. For the suprapostural task (ring task), ring-touch time (%) was calculated as the percentage of time in which the rings were touching in a walking trial.

At the cortical level, the EEG data were first filtered between 4 and 35 Hz using a zero-phase finite impulse response filter (60 dB/octave) to remove the DC shift and potential contamination from low-frequency movement artifacts < 4 Hz. A zero-phase finite impulse response band-pass filter (60 dB/octave) was selected because the filter setting is phase-invariant (FIR filter) with a sufficiently high slope to exclude EEG signals outside the range of interest (< 4 Hz and > 35 Hz) [41]. Local cortical activity and inter-regional connectivity at theta (4–7 Hz), alpha (8–12 Hz), and beta (13–35 Hz) were characterized. EEG data > 35 Hz (e.g., the gamma band) were not analyzed because of possible contamination from muscle activity of the head and neck [42]. The eye movement or blink artifacts were removed from the EEG using regression analysis with the NeuroScan 4.3 software program (NeuroScan Inc., El Paso, TX, USA), based on bipolar vertical and horizontal electrooculogram channels [43, 44]. Afterwards, the conditioned EEG data were segmented into epochs of a gait cycle (roughly 1–2 s). The EEG epochs were time-locked to single gait cycles, specifically from left heel strike to left heel strike or right heel strike to right heel strike, depending on the foot of the first heel strike recorded by the GAITRite electronic walkway. An experienced researcher visually inspected each epoch of a gait cycle to reject any remaining artifacts.

In the present study, each test trial contained 4–6 artifact-free epochs (32–48 epochs per condition). The spectral power of an epoch was calculated based on Fast Fourier Transformation with a frequency resolution of 0.2 Hz. The relative power of the epoch was computed as the ratio of the power of each spectral value to the total power, and relative power spectra were averaged across epochs for all channels. The mean value of relative power in each sub-band was obtained from the experimental trials in the GP and MP conditions. For each sub-band EEG, the inter-regional EEG connectivity of the 30 electrode pairs was estimated with the weighted phase-lag index (wPLI). This index is an extended version of the phase-lag index that maximizes the weight of±90-degree phase differences while uniformly driven (such as volume conduction) sources are effectively suppressed [45]. wPLI has been recommended for measuring EEG activity during locomotion because of its immunity to movement artifacts [46]. The wPLI weights each phase difference according to the magnitude of the lag, mathematically formulated as:

A square 30×30 wPLI-adjacent matrix of each epoch in an experimental trial for the GP and MP conditions was obtained and averaged. The wPLI of EEG electrode pairs was calculated with the HERMES function in Matlab [47]. A paired t-test was used to examine differences in the wPLI between the electrode pairs (a cell within a 30×30 matrix) between the GP and MP conditions. The resulting 30×30 t matrix represented prioritization-related differences in inter-regional connectivity between the GP and MP conditions. Supra-threshold connectivity was denoted as a high prioritization-related difference in wPLI for the two prioritization conditions (|t₃₃| > 2.733, p < 0.005). Supra-threshold connectivity was used to highlight topological differences in EEG networks between the GP and MP conditions.

Statistical analysis

Repeated two-way analysis of variance was used to compare the amount of attention directed towards the walking task and ring task between the GP and MP conditions. Paired t-tests were used to examine the effect of prioritization strategy on behavior performance and EEG metrics between the two conditions, including all gait parameters, ring-touch time, and mean relative power of the electrodes in the regions of interest. For EEG relative power, if the number of regions of interest was more than one, the Bonferroni correction would be used for multiple comparisons.

To compare differences in sub-band EEG network properties, a set of prioritization-dependent supra-threshold connectivities was mapped on the scalp. A permutation test was performed 5000 times to examine variations in the null distribution of the supra-threshold connectivity between the GP and MP conditions. Methodological details of corrected network-based statistics with the permutation test have been documented in the work of Zalesky et al. [48]. In case of a prioritization-related difference in the sub-band EEG network, mean wPLI (m-wPLI) with supra-threshold connectivity was obtained for the GP and MP conditions, and the mean strength of wPLI of the two prioritization conditions was compared with the paired t-test. To bridge behavior and EEG linkage with prioritization changes, we calculated the normalized difference (ND) in gait, ring touch, relative power of EEG electrodes in the regions of interest, and m-wPLI between the GP and MP conditions ( ND=GP-MPMP×100% ). Pearson’s correlation was used to examine the significance of the relationship between behavior variables (gait parameters and ring-touch time) and sub-band EEG metrics (relative power and m-wPLI) that exhibited significant prioritization-dependent differences. The level of significance was set at p < 0.05. Signal processing and statistical analyses were completed using MATLAB v. R2018a (MathWorks, Natick, MA, USA) and SPSS v. 21 (SPSS Inc., Chicago, IL, USA), respectively. All data are presented as means±standard deviations.

Task prioritization effect on theta power

During dual-task walking in this study, the relative theta power differed between prioritization strategies, with lower power observed in the prefrontal-frontal-center area in the GP compared with the MP condition (Fig. 2). Prefrontal/frontal theta oscillation reflects the demands of concurrent task scheduling and sustained attention [49, 50], and prefrontal/frontal theta power amplitude can be used as an indicator of cognitive load during dual-task interference. For example, compared with walking alone, walking while engaging in conversation leads to increased theta power in the frontal region [51]. In addition, increased relative theta power in the frontal or fronto-parietal area has been observed in high-demand concurrent tasks during dual-task walking or when processing high conflict information [34, 50]. Accordingly, the observed decrease in frontal theta power with the GP strategy suggests a lower cognitive load for people with PD [49]. Thus, adopting a GP strategy might effectively enhance gait performance in terms of increased walking speed and longer step length (Table 2) by promoting superior neuralefficiency.

Our findings of decreased theta power in the prefrontal-frontal and sensorimotor areas during dual-task walking with a concurrent manual task are consistent with those of Zhang et al. [29]. Using functional near infrared spectroscopy, their group also observed reduced activity in the frontal and parietal areas of individuals with PD prioritizing gait over cognition during dual-task walking. The frontocentral regions, particularly the supplementary motor area, is the primary area for regulating cognitive-motor interactions and supporting action plans for complex movements [52]. The observed decrease in sensorimotor theta power during dual-task walking with the GP strategy theoretically aligns with the extended Bland’s theta sensorimotor integration model [53]. This concordance suggests that the GP strategy reduces the cognitive effort required for movement planning and execution, possibly because of a high level of sensorimotor integration [54].

Task prioritization modulation and walking association in alpha oscillation

During dual-task walking with the GP strategy, individuals with PD showed a widespread decrease in alpha power across the frontal, sensorimotor, temporal, and occipital areas (Fig. 2). This reduction in regional alpha activity is likely associated with functional benefits for locomotion. A study investigating the effects of external auditory guidance on gait in people with PD found that during guided walking with a metronome, alpha oscillations in the parieto-occipital area were lower compared with uncued walking. This decrease in alpha activity was associated with subjective gait improvement [31]. Whittier et al. also found that increasing stride rate during running in individuals with a history of tibial injury led to suppression of alpha power in the parietal cortex [55]. In another study, training healthy young adults to decrease central alpha power using EEG neurofeedback resulted in improved performance on both the single walking task and cognitive dual-task walking [56]. In keeping with the cortical idling hypothesis [57] and gating by inhibition hypothesis [58, 59], the decreased alpha synchronization observed in the GP condition over the posterior cortical area of individuals with PD suggests an increase in conscious control and a reduction in task-switching processes during dual-task walking [60, 61]. When individuals adopt an optimal strategy, they may effectively manage the demands of both walking and manual tasks by reducing interference, inhibiting unnecessary or noisy processes, and facilitating well-integrated information processing [62]. The significant negative correlation between gait performance and alpha power in the posterior cortex (Fig. 4a, b) further supports the interpretation that alpha suppression with the GP strategy reflects enhanced attentive control over locomotion. This enhanced control leads to improve gait performance, including faster gait speed and longer step length, during dual-task walking in individuals with PD.

Task prioritization modulation and ring-task association in beta oscillation

In the GP condition, beta band activity in the left fronto-temporal area was less than in the MP condition (Fig. 2), which is in keeping with the left hemisphere’s involvement in skilled movement [63]. Deficits in force coordination of the bimanual force-matching task have been observed when the fronto-temporal cortex is excised, indicating its role in monitoring proprioceptive feedback and programming the appropriate amount of force output [64]. Individuals with early-stage PD often exhibit greater cortical atrophy of the left hemisphere, particularly in the left frontal-temporal region [63, 65]. Greater EEG beta power in the left fronto-temporal area reflects sustained attention required for response accuracy [66]. Beta oscillation in the fronto-temporal area also is associated with conflict resolution for goal-directed voluntary actions in people with PD and in unaffected adults [67]. In a hand-controlled joystick task, trials with larger tracking errors were associated with smaller beta power [68]. Of interest, with the GP strategy in the current study, participants who had smaller beta power exhibited a longer ring-touch time (Fig. 4c). A decrease in beta power at the specific site in the GP condition could represent an adaptive action selection due to gait prioritization, as it led the brain to compromise and dedicate less attentional resources to executing a concurrent manual task.

The GP strategy, however, also resulted in enhancing beta connectivity strength between the midline electrodes of Fz, FCz, Cz, CPz, and Pz to T6 (Fig. 3). This increase in circuit-specific functional connectivity due to gait prioritization may have led to an increase in ring-touch time (or poorer performance) for the manual task (Fig. 4d). To our knowledge, no previous study has reported a similar finding for dual-task situations. Our observation of greater beta oscillations suggests a potential role in maintaining the current sensorimotor set during dual-task walking with GP [69]. The enhanced beta connectivity observed in the GP condition may indicate top-down processing aimed at maintaining the coordination of big steps with the current brain state. The shift of attention away from monitoring manual performance in favor of prioritizing gait led to longer ring-touch time. The involvement of cortical midline structures in this process is believed to be related to error detection, particularly in situations with high response-conflict [70, 71]. On the other hand, the right temporo-parietal junction (T6) is responsible for spatial selective attention and self-body location in space, according to the body schema theory of the brain [72–74]. The positive relationship between ring-touch time and beta connectivity suggests that individuals with PD who devote more cognitive process to controlling locomotion in the GP condition may exhibit poorer manual performance, potentially because of reduced attention allocated to the manual task. In such a scenario, attentional resources would be directed more towards gait, leading to a trade-off in manual performance.

Overall, the decline in left regional beta activity and enhancement of the right inter-regional beta connectivity may help prevent excessive cortical resource recruitment during dual-task walking with GP. Fortunately, in behavior contexts, GP did not lead to a significant increase in ring-touch time (Table 2), despite a potential negative impact on the manual task.

Methodological concerns and study limitations

First, we analyzed power and weighted phase-lag index (wPLI) to represent average cortical activation and connectivity within a gait cycle during dual-task walking. However, it was not possible to uncover phase-dependent variations in EEG-EEG connectivity over a short time within a gait cycle. The sliding window method could be a potential solution; however, there is no consensus on the window length in the literature [75, 76], especially given its limitation in achieving high temporal resolution metrics of functional connectivity within a short gait cycle. Although this study simply assessed average connectivity and regional power within a gait cycle, we successfully demonstrated the prioritization effect on EEG dynamics for people with PD during dual-task walking. Future studies could explore variations in the prioritization effect on EEG dynamics within a gait cycle during dual-task walking. Next, the prioritization effect in PD might involve subcortical structures such as the striatum and hippocampus [77]. The activities of subcortical structures cannot be adequately characterized with scalp EEG. Third, this study lacked an age-matched control group, limiting the comparison of task prioritization effects on dual-task walking between individuals with and without PD. However, examining the prioritization of gait in healthy adults is of limited clinical significance, as they can well accomplish dual-task walking without obvious functional constraints and safety concerns. Clinically, only people with PD are necessarily benefited from a gait prioritization strategy, as reported in this study. Finally, considering different brain regions involved and the variability on dual-task paradigms, it would be prudent to extend our conclusions on prioritization effects to dual-task walking with concurrent cognitive tasks or in PD with different executive dysfunction or gait disorder (e.g., freezers vs. non-freezers) [78, 79]. The use of a dual-task paradigm involving a concurrent motor task during walking in this study is to emphasize set-shifting for the coordination of physical movements, which is believed to introduce greater task interference in limited motor resources for people with PD [23]. Further research is needed to explore the task-dependent prioritization effects on people with PD.